In [3]:
# %load jupyter_default.py
import pandas as pd
import numpy as np
import os
import re
import datetime
import time
import glob
import json
from tqdm import tqdm_notebook
from colorama import Fore, Style

%matplotlib inline
import matplotlib.pyplot as plt
import matplotlib.colors
import seaborn as sns

%config InlineBackend.figure_format='retina'
sns.set() # Revert to matplotlib defaults
plt.rcParams['figure.figsize'] = (12, 8)
plt.rcParams['axes.labelpad'] = 20
plt.rcParams['legend.fancybox'] = True
plt.style.use('ggplot')

SMALL_SIZE, MEDIUM_SIZE, BIGGER_SIZE = 14, 16, 20
plt.rc('font', size=SMALL_SIZE)
plt.rc('axes', titlesize=SMALL_SIZE)
plt.rc('axes', labelsize=MEDIUM_SIZE)
plt.rc('xtick', labelsize=SMALL_SIZE)
plt.rc('ytick', labelsize=SMALL_SIZE)
plt.rc('legend', fontsize=MEDIUM_SIZE)
plt.rc('axes', titlesize=BIGGER_SIZE)

def savefig(name):
    f_name = '../../results/figures/{}.png'.format(name)
    plt.savefig(f_name, bbox_inches='tight', dpi=300)
    print('Saving figure to {}'.format(f_name))

%reload_ext autoreload
%autoreload 2
    
%reload_ext version_information
%version_information pandas, numpy
Out[3]:
SoftwareVersion
Python3.6.8 64bit [GCC 4.2.1 Compatible Clang 4.0.1 (tags/RELEASE_401/final)]
IPython6.4.0
OSDarwin 16.7.0 x86_64 i386 64bit
pandas0.23.4
numpy1.16.3
Thu Jun 13 11:55:11 2019 PDT
In [4]:
from slugify import slugify
In [5]:
from dotenv import load_dotenv
load_dotenv('../../.env')
Out[5]:
True

Sales Forecasting

Alex's development notebook for sales forecasting.

Read data

Read from BigQuery

Run this cell to read from BigQuery.

In [49]:
%%time
"""
Using bigquery
"""

from google.cloud import bigquery
# Using environment variables to authenticate
client = bigquery.Client()

def pull_daily_data(verbose=False):
    dataset = client.get_dataset('bigquery-public-data.google_analytics_sample')

    data = []
    for table in tqdm_notebook(list(client.list_tables(dataset))):
        if verbose:
            print('Querying {}'.format(table.table_id))
        query_job = client.query('''
            select
              date,
              sum(totals.visits),
              sum(totals.pageviews),
              sum(totals.transactions),
              sum(totals.transactionRevenue)
            from `bigquery-public-data.google_analytics_sample.{}`
            group by date;
        '''.format(table.table_id))
        results = query_job.result().to_dataframe()
        results.columns = ['date', 'visits', 'pageviews', 'transactions', 'transactionRevenue']
        data.append(results)

    df = pd.concat(data, ignore_index=True, sort=False)
    return df

bq_results = pull_daily_data()

CPU times: user 7.94 s, sys: 856 ms, total: 8.8 s
Wall time: 11min 29s
In [53]:
bq_results.head()
Out[53]:
date visits pageviews transactions transactionRevenue
0 20160801 1711 9843 34 6116060000
1 20160802 2140 11784 18 1361190000
2 20160803 2890 13724 None None
3 20160804 3161 13326 17 1182890000
4 20160805 2702 13585 42 5594260000

Read jsonl from local

Run this cell if you have exported the data from BigQuery into JSON line (JSONL) files.

In [17]:
%%time
"""
Using local jsonl
"""
ERRORS = []

def pull_daily_data(verbose=False, raise_errors=False):
    if os.getenv('DATA_PATH') is None:
        raise ValueError('Please set environment variable DATA_PATH')

    dataset = sorted(glob.glob(os.path.join(os.getenv('DATA_PATH'), 'raw', '*.jsonl')))
    print('Loading from {}, etc...'.format(', '.join(dataset[:3])))

    data = []
    for table in tqdm_notebook(dataset):
        if verbose:
            print('Scanning {}'.format(table))
        with open(table, 'r') as f:
            table_data = []
            for line in f:
                d = json.loads(line)
                date = d['date']
                d = d['totals']
                try:
                    table_data.append([
                        date,
                        d['visits'],
                        d['pageviews'],
                        d['transactions'],
                        d['transactionRevenue'],
                    ])
                except Exception as e:
                    if verbose:
                        print('Error raised when reading row:\n{}'.format(e))
                    ERRORS.append([table, e])
                    if raise_errors:
                        raise(e)

            cols = ['date', 'visits', 'pageviews', 'transactions', 'transactionRevenue']
            results = (
                pd.DataFrame(table_data, columns=cols)
                    .groupby('date')[['visits', 'pageviews', 'transactions', 'transactionRevenue']]
                    .sum().reset_index()
            )
            data.append(results)

    df = pd.concat(data, ignore_index=True, sort=False)
    return df

jsonl_results = pull_daily_data()

Loading from /Volumes/GoogleDrive/My Drive/ga-data-mining/data/raw/ga_sessions_20160801.jsonl, /Volumes/GoogleDrive/My Drive/ga-data-mining/data/raw/ga_sessions_20160802.jsonl, /Volumes/GoogleDrive/My Drive/ga-data-mining/data/raw/ga_sessions_20160803.jsonl, etc...

CPU times: user 7min 50s, sys: 2min 9s, total: 10min
Wall time: 21min 21s
In [70]:
df = jsonl_results.copy()
df.date = pd.to_datetime(df.date)

f_path = os.path.join(os.getenv('DATA_PATH'), 'interim', 'sales_forecast_raw.csv')
if os.path.isfile(f_path):
    raise Exception(
        'File exists! Run line below in separate cell to overwrite it. '
        'Otherwise just run cell below to load file.')

df.to_csv(f_path, index=False)

---------------------------------------------------------------------------
Exception                                 Traceback (most recent call last)
<ipython-input-70-e5321a2c5820> in <module>()
      4 if os.path.isfile('../../data/interim/sales_forecast_raw.csv'):
      5     raise Exception(
----> 6         'File exists! Run line below in separate cell to overwrite it. '
      7         'Otherwise just run cell below to load file.')
      8 

Exception: File exists! Run line below in separate cell to overwrite it. Otherwise just run cell below to load file.

Load pre-queried data

In [7]:
def load_file(f_path):
    if not os.path.exists(f_path):
        print('No data found. Run data load script above.')
        return
    print('Loading {}'.format(f_path))
    df = pd.read_csv(f_path)
    df.date = pd.to_datetime(df.date)
    return df

f_path = os.path.join(os.getenv('DATA_PATH'), 'interim', 'sales_forecast_raw.csv')
df = load_file(f_path)

Loading /Volumes/GoogleDrive/My Drive/ga-data-mining/data/interim/sales_forecast_raw.csv
In [8]:
df.head()
Out[8]:
date visits pageviews transactions transactionRevenue
0 2016-08-01 1711 9843.0 34.0 6.116060e+09
1 2016-08-02 2140 11784.0 18.0 1.361190e+09
2 2016-08-03 2890 13724.0 NaN NaN
3 2016-08-04 3161 13326.0 17.0 1.182890e+09
4 2016-08-05 2702 13585.0 42.0 5.594260e+09
In [9]:
df['week'] = df.date.apply(lambda x: x.strftime('%W'))
df['year'] = df.date.apply(lambda x: x.strftime('%Y'))
df['week_start'] = df[['week', 'year']].apply(
    lambda x: datetime.datetime.strptime('{}-{}-1'.format(x.year, x.week), '%Y-%W-%w'),
    axis=1
)
In [10]:
df.dtypes
Out[10]:
date                  datetime64[ns]
visits                         int64
pageviews                    float64
transactions                 float64
transactionRevenue           float64
week                          object
year                          object
week_start            datetime64[ns]
dtype: object

How does the data look? Is there seasonality that I can predict?

Note: we'll throw out the first and last week (for these charts)

In [24]:
df_ = df[(df.week_start > df.week_start.min()) & (df.week_start < df.week_start.max())].copy()
In [28]:
df_.groupby('week_start').visits.sum().plot()
Out[28]:
<matplotlib.axes._subplots.AxesSubplot at 0x1a19ffd710>
In [29]:
df_.groupby('week_start').transactions.sum().plot()
Out[29]:
<matplotlib.axes._subplots.AxesSubplot at 0x1a1a02cfd0>

Next we'll try to forecast this trend forward. It will be difficult, since there's only one year of data to train models on.

Forcasting with Facebook Prophet

In [11]:
from fbprophet import Prophet
import warnings
# Ignore warnings from prophet lib
warnings.filterwarnings('ignore', 'Conversion of the second argument of issubdtype')

Transactions

Making sure to add yearly_seasonality as we expect yearly periodicity e.g. Christmas, black friday, etc...

Tune the parameter values (weekly_seasonality and yearly_seasonality and seasonality_prior_scale) to fit the given data. Higher values will lead to tighter fits (be careful about overfitting here).

In [52]:
df_prophet = df[['date', 'transactions']]\
    .rename(columns={'date': 'ds', 'transactions': 'y'})
df_prophet['ds'] = df_prophet['ds'].apply(lambda x: x.strftime('%Y-%m-%d'))

m = Prophet(
    daily_seasonality=False,
    weekly_seasonality=1,
    yearly_seasonality=20,
    seasonality_prior_scale=0.1,
)
m.fit(df_prophet)

future = m.make_future_dataframe(periods=365, freq='D')
forecast = m.predict(future)
fig = m.plot(forecast)
plt.ylabel('Transactions')
plt.xlabel('Date')
savefig('sales_forecast_transactions')

/anaconda3/lib/python3.6/site-packages/pystan/misc.py:399: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
  elif np.issubdtype(np.asarray(v).dtype, float):
In [53]:
fig = m.plot_components(forecast)
fig.axes[0].set_xlabel('Date')
fig.axes[0].set_ylabel('Overall Trend')
fig.axes[1].set_ylabel('Weekly Trend')
fig.delaxes(fig.axes[2])
savefig('sales_forecast_transactions_trends')

Sales

In [14]:
df_prophet = df[['date', 'transactionRevenue']]\
    .rename(columns={'date': 'ds', 'transactionRevenue': 'y'})
df_prophet['y'] = df_prophet['y'] / 1e6
df_prophet['ds'] = df_prophet['ds'].apply(lambda x: x.strftime('%Y-%m-%d'))

m = Prophet(
    daily_seasonality=False,
    weekly_seasonality=1,
    yearly_seasonality=20,
    seasonality_prior_scale=0.1,
)
m.fit(df_prophet)

future = m.make_future_dataframe(periods=365, freq='D')
forecast = m.predict(future)
fig = m.plot(forecast)
plt.ylabel('Sales (USD)')
plt.xlabel('Date')

savefig('sales_forecast')

Saving figure to ../../results/figures/sales_forecast.png
In [15]:
fig = m.plot_components(forecast)
fig.axes[0].set_xlabel('Date')
fig.axes[0].set_ylabel('Overall Trend')
fig.axes[1].set_ylabel('Weekly Trend')
fig.delaxes(fig.axes[2])
savefig('sales_forecast_trends')

Saving figure to ../../results/figures/sales_forecast_trends.png

Predictions by quarter

In [55]:
forecast['date'] = pd.to_datetime(forecast['ds'])
In [29]:
def add_quarters(df):
    """
    Add labels for quarter number. May need to adjust for specific
    fiscal calendar. Only covers until end of 2018.
    
    df : pd.DataFrame
        Should have "date" column that is datetime dtype.
    """
    df['quarter'] = float('nan')
    df['quarter_num'] = float('nan')
    
    df.loc[(df.date >= datetime.datetime(2016, 7, 1))&(df.date < datetime.datetime(2016, 10, 1)), 'quarter'] = '2016 Q3'
    df.loc[(df.date >= datetime.datetime(2016, 10, 1))&(df.date < datetime.datetime(2017, 1, 1)), 'quarter'] = '2016 Q4'
    df.loc[(df.date >= datetime.datetime(2017, 1, 1))&(df.date < datetime.datetime(2017, 4, 1)), 'quarter'] = '2017 Q1'
    df.loc[(df.date >= datetime.datetime(2017, 4, 1))&(df.date < datetime.datetime(2017, 7, 1)), 'quarter'] = '2017 Q2'
    df.loc[(df.date >= datetime.datetime(2017, 7, 1))&(df.date < datetime.datetime(2017, 10, 1)), 'quarter'] = '2017 Q3'
    df.loc[(df.date >= datetime.datetime(2017, 10, 1))&(df.date < datetime.datetime(2018, 1, 1)), 'quarter'] = '2017 Q4'
    df.loc[(df.date >= datetime.datetime(2018, 1, 1))&(df.date < datetime.datetime(2018, 4, 1)), 'quarter'] = '2018 Q1'
    df.loc[(df.date >= datetime.datetime(2018, 4, 1))&(df.date < datetime.datetime(2018, 7, 1)), 'quarter'] = '2018 Q2'
    df.loc[(df.date >= datetime.datetime(2018, 7, 1))&(df.date < datetime.datetime(2018, 10, 1)), 'quarter'] = '2018 Q3'
    df.loc[(df.date >= datetime.datetime(2018, 10, 1))&(df.date < datetime.datetime(2019, 1, 1)), 'quarter'] = '2018 Q4'

    df.loc[(df.date >= datetime.datetime(2016, 7, 1))&(df.date < datetime.datetime(2016, 10, 1)), 'quarter_num'] = 1
    df.loc[(df.date >= datetime.datetime(2016, 10, 1))&(df.date < datetime.datetime(2017, 1, 1)), 'quarter_num'] = 2
    df.loc[(df.date >= datetime.datetime(2017, 1, 1))&(df.date < datetime.datetime(2017, 4, 1)), 'quarter_num'] = 3
    df.loc[(df.date >= datetime.datetime(2017, 4, 1))&(df.date < datetime.datetime(2017, 7, 1)), 'quarter_num'] = 4
    df.loc[(df.date >= datetime.datetime(2017, 7, 1))&(df.date < datetime.datetime(2017, 10, 1)), 'quarter_num'] = 5
    df.loc[(df.date >= datetime.datetime(2017, 10, 1))&(df.date < datetime.datetime(2018, 1, 1)), 'quarter_num'] = 6
    df.loc[(df.date >= datetime.datetime(2018, 1, 1))&(df.date < datetime.datetime(2018, 4, 1)), 'quarter_num'] = 7
    df.loc[(df.date >= datetime.datetime(2018, 4, 1))&(df.date < datetime.datetime(2018, 7, 1)), 'quarter_num'] = 8
    df.loc[(df.date >= datetime.datetime(2018, 7, 1))&(df.date < datetime.datetime(2018, 10, 1)), 'quarter_num'] = 9
    df.loc[(df.date >= datetime.datetime(2018, 10, 1))&(df.date < datetime.datetime(2019, 1, 1)), 'quarter_num'] = 10
    return df
In [58]:
forecast = add_quarters(forecast)
In [59]:
forecast.quarter.value_counts()
Out[59]:
2016 Q4    92
2017 Q3    92
2017 Q4    92
2018 Q2    91
2017 Q2    91
2018 Q1    90
2017 Q1    90
2016 Q3    61
2018 Q3    32
Name: quarter, dtype: int64

For the Google Merch Store, we only have partial data for 2016 & 2018 Q3, so we'll want to filter these out.

In [62]:
m_google_merch = (forecast.quarter != '2016 Q3') & (forecast.quarter != '2018 Q3')

s_transactionRevenue = (
    forecast[m_google_merch].groupby(['quarter_num', 'quarter'])
        .yhat.sum().reset_index()
        .set_index('quarter_num').sort_index(ascending=True)
        .set_index('quarter')['yhat']
)

# Print it
s_transactionRevenue.apply(lambda x: '${:,}'.format(round(x)))
Out[62]:
quarter
2016 Q4    $385,919
2017 Q1    $341,650
2017 Q2    $399,708
2017 Q3    $486,170
2017 Q4    $495,440
2018 Q1    $447,201
2018 Q2    $508,503
Name: yhat, dtype: object

Display these results in a dataframe

In [63]:
# Get the actual sales (as opposed to predicted above)

df = add_quarters(df)
m_google_merch = (df.quarter != '2016 Q3') & (df.quarter != '2018 Q3')
s_actual_transactionRevenue = (
    df[m_google_merch].groupby(['quarter_num', 'quarter'])
        .transactionRevenue.sum().reset_index()
        .set_index('quarter_num').sort_index(ascending=True)
        .set_index('quarter')['transactionRevenue'] / 1e6
)

# Print it
s_actual_transactionRevenue.apply(lambda x: '${:,}'.format(round(x)))
Out[63]:
quarter
2016 Q4    $386,901
2017 Q1    $337,599
2017 Q2    $402,070
2017 Q3    $132,804
Name: transactionRevenue, dtype: object

Generate a YoY table view

In [67]:
forecast_results = pd.DataFrame({
    'Reporting Period': ['Q4', 'Q1', 'Q2'],
    'Prev Year': [
        s_actual_transactionRevenue[s_actual_transactionRevenue.index=='2016 Q4'].values[0] if (s_actual_transactionRevenue.index=='2016 Q4').sum() else 0,
        s_actual_transactionRevenue[s_actual_transactionRevenue.index=='2017 Q1'].values[0] if (s_actual_transactionRevenue.index=='2017 Q1').sum() else 0,
        s_actual_transactionRevenue[s_actual_transactionRevenue.index=='2017 Q2'].values[0] if (s_actual_transactionRevenue.index=='2017 Q2').sum() else 0,
    ],
    'Forecasted': [
        s_transactionRevenue[s_transactionRevenue.index=='2017 Q4'].values[0],
        s_transactionRevenue[s_transactionRevenue.index=='2018 Q1'].values[0],
        s_transactionRevenue[s_transactionRevenue.index=='2018 Q2'].values[0],
    ],
})
forecast_results['YoY (%)'] = ((forecast_results['Forecasted'] - forecast_results['Prev Year'])
                                / forecast_results['Prev Year'] * 100).apply(lambda x: '{:+.0f}%'.format(x))
forecast_results['Prev Year'] = forecast_results['Prev Year'].apply(lambda x: '${:,.0f}'.format(x))
forecast_results['Forecasted'] = forecast_results['Forecasted'].apply(lambda x: '${:,.0f}'.format(x))
forecast_results.set_index('Reporting Period', inplace=True)
forecast_results.to_csv('../../results/tables/sales_forecast.csv')
In [68]:
forecast_results
Out[68]:
Prev Year Forecasted YoY (%)
Reporting Period
Q4 $386,901 $495,440 +28%
Q1 $337,599 $447,201 +32%
Q2 $402,070 $508,503 +26%

Forecasting by device, region, source, etc...

The Google Merch Store dataset in BigQuery gives limited access to segmentation variables. Some interesting options we do have access to are device, region and traffic source.

Read from BigQuery

Run this cell to read from BigQuery.

In [215]:
%%time
"""
Using bigquery
"""

from google.cloud import bigquery
# Using environment variables to authenticate
client = bigquery.Client()

def pull_daily_data(verbose=False):
    dataset = client.get_dataset('bigquery-public-data.google_analytics_sample')

    data = []
    for table in tqdm_notebook(list(client.list_tables(dataset))):
        if verbose:
            print('Querying {}'.format(table.table_id))
        query_job = client.query('''
            select
              date,
              sum(totals.visits),
              sum(totals.pageviews),
              sum(totals.transactions),
              sum(totals.transactionRevenue),
              trafficSource.source as source,
              device.deviceCategory as deviceCategory,
              geoNetwork.country as country,
              geoNetwork.region as region
            from `bigquery-public-data.google_analytics_sample.{}`
            group by date, source, deviceCategory, country, region;
        '''.format(table.table_id))
        results = query_job.result().to_dataframe()
        results.columns = ['date', 'visits', 'pageviews', 'transactions', 'transactionRevenue',
                           'source', 'deviceCategory', 'country', 'region']
        data.append(results)

    df = pd.concat(data, ignore_index=True, sort=False)
    return df

bq_results = pull_daily_data()

CPU times: user 14.2 s, sys: 1.86 s, total: 16.1 s
Wall time: 13min 31s
In [216]:
bq_results.head()
Out[216]:
date visits pageviews transactions transactionRevenue source deviceCategory country region
0 20160801 3 4.0 NaN NaN youtube.com desktop Macedonia (FYROM) not available in demo dataset
1 20160801 1 2.0 NaN NaN squishable.com mobile Japan not available in demo dataset
2 20160801 1 1.0 NaN NaN youtube.com desktop Cyprus (not set)
3 20160801 3 3.0 NaN NaN (direct) desktop Singapore not available in demo dataset
4 20160801 1 1.0 NaN NaN (direct) mobile China not available in demo dataset
In [217]:
df = bq_results.copy()
df.date = pd.to_datetime(df.date)

f_path = os.path.join(os.getenv('DATA_PATH'), 'interim', 'sales_forecast_segments_raw.csv')
if os.path.isfile(f_path):
    raise Exception(
        'File exists! Run line below in separate cell to overwrite it. '
        'Otherwise just run cell below to load file.')

df.to_csv(f_path, index=False)

Load pre-queried data

In [16]:
def load_file(f_path):
    if not os.path.exists(f_path):
        raise Exception('No data found. Run data load script above.')
    print('Loading {}'.format(f_path))
    df = pd.read_csv(f_path)
    df.date = pd.to_datetime(df.date)
    return df

f_path = os.path.join(os.getenv('DATA_PATH'), 'interim', 'sales_forecast_segments_raw.csv')
df = load_file(f_path)

Loading /Volumes/GoogleDrive/My Drive/ga-data-mining/data/interim/sales_forecast_segments_raw.csv
In [17]:
df.head()
Out[17]:
date visits pageviews transactions transactionRevenue source deviceCategory country region
0 2016-08-01 3 4.0 NaN NaN youtube.com desktop Macedonia (FYROM) not available in demo dataset
1 2016-08-01 1 2.0 NaN NaN squishable.com mobile Japan not available in demo dataset
2 2016-08-01 1 1.0 NaN NaN youtube.com desktop Cyprus (not set)
3 2016-08-01 3 3.0 NaN NaN (direct) desktop Singapore not available in demo dataset
4 2016-08-01 1 1.0 NaN NaN (direct) mobile China not available in demo dataset
In [18]:
df['week'] = df.date.apply(lambda x: x.strftime('%W'))
df['year'] = df.date.apply(lambda x: x.strftime('%Y'))
df['week_start'] = df[['week', 'year']].apply(
    lambda x: datetime.datetime.strptime('{}-{}-1'.format(x.year, x.week), '%Y-%W-%w'),
    axis=1
)
In [19]:
df.dtypes
Out[19]:
date                  datetime64[ns]
visits                         int64
pageviews                    float64
transactions                 float64
transactionRevenue           float64
source                        object
deviceCategory                object
country                       object
region                        object
week                          object
year                          object
week_start            datetime64[ns]
dtype: object
In [20]:
df_ = df[(df.week_start > df.week_start.min()) & (df.week_start < df.week_start.max())].copy()
In [224]:
df_.groupby('week_start').visits.sum().plot()
Out[224]:
<matplotlib.axes._subplots.AxesSubplot at 0x1a37bfe2b0>
In [225]:
df_.groupby('week_start').transactions.sum().plot()
Out[225]:
<matplotlib.axes._subplots.AxesSubplot at 0x1a2be76588>

Next we'll try to forecast this trend forward. It will be difficult, since there's only one year of data to train models on.

Forcasting with Transactions

In [226]:
from fbprophet import Prophet
import warnings
# Ignore warnings from prophet lib
warnings.filterwarnings('ignore', 'Conversion of the second argument of issubdtype')

Making sure to add yearly_seasonality as we expect yearly periodicity e.g. Christmas, black friday, etc...

Tune the parameter values (weekly_seasonality and yearly_seasonality and seasonality_prior_scale) to fit the given data. Higher values will lead to tighter fits (be careful about overfitting here).

In [244]:
Prophet.plot?
In [21]:
def build_prophet_df(df, segment_col):
    df_prophet = (
        df.groupby(['date', segment_col]).transactions.sum()
        .reset_index()
        .rename(columns={'transactions': 'y'})
    )
    df_prophet['ds'] = df_prophet['date'].apply(lambda x: x.strftime('%Y-%m-%d'))
    df_prophet = df_prophet.sort_values('date', ascending=True)
    return df_prophet

def segment_forecast_daily(
    df,
    segment_col,
    max_num_segments=6,
    daily_seasonality=False,
    weekly_seasonality=2,
    yearly_seasonality=10,
    seasonality_prior_scale=0.1,
):
    df_prophet = build_prophet_df(df, segment_col)
    
    segs = (
        df_prophet.groupby(segment_col)['y'].sum()
        .sort_values(ascending=False)
        .head(max_num_segments).index.tolist()
    )
    models = []
    forecasts = []
    for seg in segs:
        m = Prophet(
            daily_seasonality=daily_seasonality,
            weekly_seasonality=weekly_seasonality,
            yearly_seasonality=yearly_seasonality,
            seasonality_prior_scale=seasonality_prior_scale,
        )
        m.fit(df_prophet[df_prophet[segment_col] == seg])
        models.append(m)
        
        future = m.make_future_dataframe(periods=365, freq='D')
        forecast = m.predict(future)
        forecasts.append(forecast)
        
    colors = list('rbgykmc') * 10
    fig, axes = plt.subplots(len(forecasts), sharex=True)
    for i, (ax, seg, forecast, c) in enumerate(zip(axes, segs, forecasts, colors)):
        forecast.set_index('ds')['yhat'].plot(label=seg, ax=ax, color=c)
        (df_prophet[df_prophet[segment_col] == seg]
         .set_index('date')['y'].plot(label='_', ax=ax, color=c, marker='o', linewidth=0))
        ax.legend()
        if i == int(len(forecasts) / 2):
            ax.set_ylabel('Transactions')
        else:
            ax.set_ylabel('')

    plt.xlabel('')
    savefig('sales_forecast_{}_transactions'.format('-'.join(segment_col.split(' '))))
    plt.show()

    for seg, m, forecast, c in zip(segs, models, forecasts, colors):
        fig = m.plot_components(forecast)
        fig.axes[0].set_xlabel('{} - Date'.format(seg.title()))
        fig.axes[0].set_ylabel('Overall Trend')
        fig.axes[0].lines[0].set_color(c)
        fig.axes[1].set_xlabel('{} - Day of week'.format(seg.title()))
        fig.axes[1].set_ylabel('Weekly Trend')
        fig.axes[1].lines[0].set_color(c)
        fig.delaxes(fig.axes[2])
        savefig('sales_forecast_{}={}_transactions_trends'.format(
            '-'.join(segment_col.split(' ')),
            '-'.join(seg.split(' '))
        ))
        plt.show()
In [330]:
segment_forecast_daily(df, 'deviceCategory')

Saving figure to ../../results/figures/sales_forecast_deviceCategory_transactions.png

Saving figure to ../../results/figures/sales_forecast_deviceCategory=desktop_transactions_trends.png

Saving figure to ../../results/figures/sales_forecast_deviceCategory=mobile_transactions_trends.png

Saving figure to ../../results/figures/sales_forecast_deviceCategory=tablet_transactions_trends.png

Cool! We can see distinct trends for day of week on mobile VS desktop.

In [22]:
df_fltr = df[~(df.source.isin(['not available in demo dataset', '(not set)']))]
segment_forecast_daily(df_fltr, 'source', max_num_segments=4)

Saving figure to ../../results/figures/sales_forecast_source_transactions.png

Saving figure to ../../results/figures/sales_forecast_source=(direct)_transactions_trends.png

Saving figure to ../../results/figures/sales_forecast_source=google_transactions_trends.png

Saving figure to ../../results/figures/sales_forecast_source=dfa_transactions_trends.png

Saving figure to ../../results/figures/sales_forecast_source=mail.google.com_transactions_trends.png
In [23]:
df_fltr = df[~(df.country.isin(['not available in demo dataset', '(not set)']))]
segment_forecast_daily(df_fltr, 'country', max_num_segments=3)

Saving figure to ../../results/figures/sales_forecast_country_transactions.png

Saving figure to ../../results/figures/sales_forecast_country=United-States_transactions_trends.png

Saving figure to ../../results/figures/sales_forecast_country=Canada_transactions_trends.png

Saving figure to ../../results/figures/sales_forecast_country=Venezuela_transactions_trends.png
In [333]:
df_fltr = df[~(df.region.isin(['not available in demo dataset', '(not set)']))]
segment_forecast_daily(df_fltr, 'region')

Saving figure to ../../results/figures/sales_forecast_region_transactions.png

Saving figure to ../../results/figures/sales_forecast_region=California_transactions_trends.png

Saving figure to ../../results/figures/sales_forecast_region=New-York_transactions_trends.png

Saving figure to ../../results/figures/sales_forecast_region=Illinois_transactions_trends.png

Saving figure to ../../results/figures/sales_forecast_region=Washington_transactions_trends.png

Saving figure to ../../results/figures/sales_forecast_region=Texas_transactions_trends.png

Saving figure to ../../results/figures/sales_forecast_region=Michigan_transactions_trends.png

Forecasting by product

Read from bigquery

In [ ]:
%%time
"""
Using bigquery
"""

def pull_daily_product_sales(verbose=False):
    dataset = client.get_dataset('bigquery-public-data.google_analytics_sample')

    data = []
    for table in tqdm_notebook(list(client.list_tables(dataset))):
        if verbose:
            print('Querying {}'.format(table.table_id))
        query_job = client.query('''
        SELECT
            h.item.productName AS other_purchased_products,
            COUNT(h.item.productName) AS quantity
        FROM `bigquery-public-data.google_analytics_sample.{}`,
            UNNEST(hits) as h
        WHERE (
            fullVisitorId IN (
                SELECT fullVisitorId
                FROM `bigquery-public-data.google_analytics_sample.{}`,
                    UNNEST(hits) as h
                WHERE h.item.productName CONTAINS 'Product Item Name A'
                AND totals.transactions>=1
                GROUP BY fullVisitorId
            )
            AND h.item.productName IS NOT NULL
            AND h.item.productName != 'Product Item Name A'
        )
        GROUP BY other_purchased_products
        ORDER BY quantity DESC;
        '''.format(table.table_id, table.table_id))
        results = query_job.result().to_dataframe()
        results.columns = ['date', 'visits', 'pageviews', 'transactions', 'transactionRevenue']
        data.append(results)

    df = pd.concat(data, ignore_index=True, sort=False)
    return df

bq_product_results = pull_daily_product_sales()
In [ ]:
bq_product_results.head()
In [ ]:
df = bq_product_results.copy()
df.date = pd.to_datetime(df.date)

f_path = os.path.join(os.getenv('DATA_PATH'), 'interim', 'sales_forecast_by_product_raw.csv')
if os.path.isfile(f_path):
    raise Exception(
        'File exists! Run line below in separate cell to overwrite it. '
        'Otherwise just run cell below to load file.')

df.to_csv(f_path, index=False)

Read from jsonl local

In [304]:
from typing import List, Tuple, Dict
In [305]:
%%time
"""
Using local jsonl
"""
ERRORS = []

def pull_daily_product_sales(
    verbose=False,
    raise_errors=False,
    test=False,
) -> Tuple[pd.DataFrame, dict]:
    if os.getenv('DATA_PATH') is None:
        raise ValueError('Please set environment variable DATA_PATH')

    dataset = sorted(glob.glob(os.path.join(os.getenv('DATA_PATH'), 'raw', '*.jsonl')))
    print('Loading from {}, etc...'.format(', '.join(dataset[:3])))

    data = []
    for table in tqdm_notebook(dataset):
        if verbose:
            print('Scanning {}'.format(table))
        with open(table, 'r') as f:
            for line in f:
                d = json.loads(line)
                try:
                    if not d['totals']['transactions']:
                        # No purchases, continue to next visitor
                        continue
                    for hit in d['hits']:
                        for product in hit['product']:
                            if product['productRevenue']:
                                data.append({
                                    'date': d['date'],
                                    'visitId': d['visitId'],
                                    'fullVisitorId': d['fullVisitorId'],
                                    'product': product,
                                })
                except Exception as e:
                    if verbose:
                        print('Error raised when reading row:\n{}'.format(e))
                    ERRORS.append([table, e])
                    if raise_errors:
                        raise(e)
                        
        if test and (table == dataset[1]):
            break

    cols_main = ['date', 'visitId', 'fullVisitorId']
    cols_product = [
        'productSKU', 'v2ProductName', 'v2ProductCategory', 'productVariant',
        'productRevenue', 'productQuantity', 'productRefundAmount'
    ]
    df_data = [
        [d.get(col, float('nan')) for col in cols_main]
        + [d['product'].get(col) for col in cols_product]
        for d in data
    ]
    df = pd.DataFrame(df_data, columns=(cols_main+cols_product))
    return df, data

jsonl_product_results, nosql_data = pull_daily_product_sales(raise_errors=True, test=True)

CPU times: user 2.74 s, sys: 105 ms, total: 2.84 s
Wall time: 2.91 s
In [306]:
jsonl_product_results, nosql_data = pull_daily_product_sales()



















In [120]:



    


df = jsonl_product_results.copy()
df.date = pd.to_datetime(df.date)

f_path = os.path.join(os.getenv('DATA_PATH'), 'interim', 'sales_forecast_by_product_raw.csv')
if os.path.isfile(f_path):
    raise Exception(
        'File exists! Run line below in separate cell to overwrite it. '
        'Otherwise just run cell below to load file.')

df.to_csv(f_path, index=False)


























---------------------------------------------------------------------------
Exception                                 Traceback (most recent call last)
<ipython-input-120-9adc11fa2226> in <module>()
      5 if os.path.isfile(f_path):
      6     raise Exception(
----> 7         'File exists! Run line below in separate cell to overwrite it. '
      8         'Otherwise just run cell below to load file.')
      9 

Exception: File exists! Run line below in separate cell to overwrite it. Otherwise just run cell below to load file.






















Load pre-queried data












In [6]:



    


def load_file(f_path):
    if not os.path.exists(f_path):
        print('No data found. Run data load script above.')
        return
    print('Loading {}'.format(f_path))
    df = pd.read_csv(f_path)
    df.date = pd.to_datetime(df.date)
    return df

f_path = os.path.join(os.getenv('DATA_PATH'), 'interim', 'sales_forecast_by_product_raw.csv')
df = load_file(f_path)


























Loading /Volumes/GoogleDrive/My Drive/ga-data-mining/data/interim/sales_forecast_by_product_raw.csv



















In [122]:



    


df.head()




















Out[122]:











  
    

      
      date
      visitId
      fullVisitorId
      productSKU
      v2ProductName
      productVariant
      productRevenue
      productQuantity
      productRefundAmount
    

  
  
    

      0
      2016-08-01
      1470106850
      4993485206334150199
      GGOEYAAQ031714
      YouTube Men's Short Sleeve Hero Tee White
      MD
      23990000
      1
      NaN
    

    

      1
      2016-08-01
      1470084579
      3907770685196037697
      GGOEYDHJ019399
      24 oz YouTube Sergeant Stripe Bottle
      Single Option Only
      12990000
      1
      NaN
    

    

      2
      2016-08-01
      1470069291
      7056870819058799369
      GGOEGAAQ010417
      Google Men's 100% Cotton Short Sleeve Hero Tee...
      2XL
      22590000
      1
      NaN
    

    

      3
      2016-08-01
      1470096817
      6030957980134486247
      GGOEYAEB028413
      Women's YouTube Short Sleeve Hero Tee Black
      SM
      19990000
      1
      NaN
    

    

      4
      2016-08-01
      1470055898
      6203401114057579951
      GGOEGDHC018299
      22 oz Mini Mountain Bottle
      BLUE
      36350000
      15
      NaN
    

  


























Exploration
















Top selling products













In [77]:



    


df.v2ProductName.value_counts(ascending=False).head(10)




















Out[77]:








Google Sunglasses                                       1364
Google Laptop and Cell Phone Stickers                    788
Google 22 oz Water Bottle                                784
Google Men's 100% Cotton Short Sleeve Hero Tee White     760
Google Men's 100% Cotton Short Sleeve Hero Tee Black     715
Google Men's 100% Cotton Short Sleeve Hero Tee Navy      561
Google Men's  Zip Hoodie                                 468
Recycled Paper Journal Set                               431
BLM Sweatshirt                                           405
Engraved Ceramic Google Mug                              400
Name: v2ProductName, dtype: int64























Top selling products by revenue













In [78]:



    


(df.groupby('v2ProductName').productRevenue.sum() / 1e6).sort_values(ascending=False).head(10)




















Out[78]:








v2ProductName
Google Men's  Zip Hoodie                                47636.271605
26 oz Double Wall Insulated Bottle                      44454.491527
Google 22 oz Water Bottle                               42995.041996
Leatherette Journal                                     38563.457656
Google Sunglasses                                       35490.321293
Google Metallic Notebook Set                            28846.596814
Google Men's 100% Cotton Short Sleeve Hero Tee Black    28047.612114
Recycled Paper Journal Set                              27918.170430
Google Men's 100% Cotton Short Sleeve Hero Tee White    27856.018891
Google Hard Cover Journal                               25432.673189
Name: productRevenue, dtype: float64























How many of the same product are ordered together?













In [16]:



    


s = df.productQuantity.value_counts().sort_index(ascending=False)
s.name = 'Number of product duplicates in basket'
pd.DataFrame(index=list(range(2,21))).join(s).iloc[::-1].plot.barh(color='b')
savefig('sales_forecast_product_duplicates')


























Saving figure to ../../results/figures/sales_forecast_product_duplicates.png









































Most common groups of 2













In [147]:



    


s = df[df.productQuantity == 2].groupby(['v2ProductName', 'productVariant']).size().sort_values(ascending=False)
s.name = 'Counts'
s.reset_index().set_index('Counts').head()




















Out[147]:











  
    

      
      v2ProductName
      productVariant
    

    

      Counts
      
      
    

  
  
    

      76
      Google Laptop and Cell Phone Stickers
      Single Option Only
    

    

      53
      Google Sunglasses
      BLUE
    

    

      51
      Engraved Ceramic Google Mug
      Single Option Only
    

    

      49
      YouTube Custom Decals
      Single Option Only
    

    

      48
      Google 22 oz Water Bottle
      BLUE
    

  


























Most common groups of 3













In [149]:



    


s = df[df.productQuantity == 3].groupby(['v2ProductName', 'productVariant']).size().sort_values(ascending=False)
s.name = 'Counts'
s.reset_index().set_index('Counts').head()




















Out[149]:











  
    

      
      v2ProductName
      productVariant
    

    

      Counts
      
      
    

  
  
    

      40
      Google Laptop and Cell Phone Stickers
      Single Option Only
    

    

      33
      Google 22 oz Water Bottle
      BLUE
    

    

      31
      Windup Android
      Single Option Only
    

    

      23
      Google Sunglasses
      BLUE
    

    

      21
      Google 22 oz Water Bottle
      RED
    

  


























How many different products are purchased together? i.e. final basket sizes













In [8]:



    


# Add a column for transaction ID
# See docs for more details on this: https://support.google.com/analytics/answer/3437719?hl=en
df['transactionId'] = df['visitId'].astype(str) + '|' + df['fullVisitorId'].astype(str)

















In [15]:



    


m = ~(df[['transactionId', 'v2ProductName', 'productVariant']].duplicated())
s = df[m].groupby('transactionId').size().value_counts().sort_index(ascending=False)
s.name = 'Number of distinct products in basket'

pd.DataFrame(index=list(range(1,21))).join(s).iloc[::-1].plot.barh(color='b')
# savefig('sales_forecast_unique_basket_products')




















Out[15]:








<matplotlib.axes._subplots.AxesSubplot at 0x11dc95ba8>









































How does this look on a % scale













In [13]:



    


from matplotlib.ticker import FuncFormatter

m = ~(df[['transactionId', 'v2ProductName', 'productVariant']].duplicated())
s = df[m].groupby('transactionId').size().value_counts().sort_index(ascending=False)
s = s / s.sum()
s.name = 'Number of distinct products in basket'

ax = pd.DataFrame(index=list(range(1,21))).join(s).iloc[::-1].plot.barh(color='b')
ax.xaxis.set_major_formatter(FuncFormatter(lambda y, _: '{:.0%}'.format(y))) 
savefig('sales_forecast_unique_basket_products')


























Saving figure to ../../results/figures/sales_forecast_unique_basket_products.png









































Which 2 items are most commonly purchased together?













In [205]:



    


fig = plt.figure(figsize=(4, 2))

m = ~(df[['transactionId', 'v2ProductName', 'productVariant']].duplicated())
s = df[m].groupby('transactionId').size()
transaction_ids = s[s == 2].index.tolist()

pairs = []
for transaction_id in tqdm_notebook(transaction_ids):
    data = df[df.transactionId == transaction_id][['v2ProductName', 'productVariant']].values.tolist()
    if len(data) != 2:
        continue
    
    pair = ', '.join(['{} ({})'.format(d[0].strip(), d[1].strip()) for d in data])
    pairs.append(pair)
    
s = pd.Series(pairs).value_counts(ascending=False)

s.head().iloc[::-1].plot.barh(color='b')

savefig('sales_forecast_product_pairs_1')


















































Same chart but ignoring product variants













In [207]:



    


fig = plt.figure(figsize=(4, 2))

m = ~(df[['transactionId', 'v2ProductName']].duplicated())
s = df[m].groupby('transactionId').size()
transaction_ids = s[s == 2].index.tolist()

pairs = []
for transaction_id in tqdm_notebook(transaction_ids):
    data = df[df.transactionId == transaction_id]['v2ProductName'].values.tolist()
    if len(data) != 2:
        continue
    
    pair = ', '.join(data)
    pairs.append(pair)
    
s = pd.Series(pairs).value_counts(ascending=False)

s.head().iloc[::-1].plot.barh(color='b')

savefig('sales_forecast_product_pairs_2')


























 
 














































Forecasts by product
















In general it's difficult to make granular predictions like this on a daily basis. For this reason we will group by week.


In order to help accurately account for anomaly events (e.g. someone places an order for 100 hoddies), I am going to remove outliers (by computing standard deviation and filtering out e.g. >2s.d.). Then I'll take that missing revenue and add it back into the quarterly predictions, splitting evenly.













In [46]:



    


def product_forcast(
    df,
    product_name,
    ignore_std=2,
    yearly_seasonality=10,
    weekly_seasonality=False,
    daily_seasonality=False,
    add_back_outlier_revenue=False,
) -> pd.DataFrame:

    # Filter on product & remove outliers
    outlier_sum = 0
    df_ = df[df.v2ProductName == product_name].copy()
    m = pd.Series(True, index=df_.index)
    if ignore_std:
        m = df_.productRevenue < (df_.productRevenue.mean() + df_.productRevenue.std()*ignore_std)
        outlier_sum = df_.loc[~m, 'productRevenue'].sum() / 1e6
        print('Ignoring {:} outlier points (${:,})'.format((~m).sum(), outlier_sum))

    if add_back_outlier_revenue:
        print('Warning, adding back outlier revenue to forecast but not '
             'including it in previous year numbers. Set '
              'add_back_outlier_revenue=False to fix this.')
    else:
        outlier_sum = 0
        
    print('Found {} product transactions'.format(m.sum()))
    if m.sum() == 0:
        print('Returning empty DataFrame')
        return pd.DataFrame()
    
    # Group by week
    df_ = (df_[m]
           .groupby(pd.Grouper(key='date', freq='W-MON'))
           .productRevenue.sum().reset_index())

    df_prophet = df_[['date', 'productRevenue']]\
        .rename(columns={'date': 'ds', 'productRevenue': 'y'})
    df_prophet['y'] = df_prophet['y'] / 1e6
    df_prophet['ds'] = df_prophet['ds'].apply(lambda x: x.strftime('%Y-%m-%d'))

    model = Prophet(yearly_seasonality=yearly_seasonality,
                weekly_seasonality=weekly_seasonality,
                daily_seasonality=daily_seasonality)
    model.fit(df_prophet)

    future = model.make_future_dataframe(periods=52, freq='7D')
    forecast = model.predict(future)
    model.plot(forecast)
    plt.ylim(bottom=0)
    plt.ylabel('{} Revenue (USD)'.format(product_name.title()))
    plt.xlabel('Date')
    savefig('sales_forecast_{}'.format(slugify(product_name)))
    
    print('Generating quarterly forecasts')
    forecast['date'] = pd.to_datetime(forecast['ds'])
    forecast = add_quarters(forecast)

    m_q = (forecast.quarter != '2016 Q3') & (forecast.quarter != '2018 Q3')
    s_transactions = (
        forecast[m_q].groupby(['quarter_num', 'quarter'])
            .yhat.sum().reset_index()
            .set_index('quarter_num').sort_index(ascending=True)
            .set_index('quarter')['yhat']
    )

    # Get the actual sales (as opposed to predicted above)
    df_ = add_quarters(df_)
    m_q = (df_.quarter != '2016 Q3') & (df_.quarter != '2018 Q3')
    s_actual_productRevenue = (
        df_[m_q].groupby(['quarter_num', 'quarter'])
            .productRevenue.sum().reset_index()
            .set_index('quarter_num').sort_index(ascending=True)
            .set_index('quarter')['productRevenue'] / 1e6
    )

    forecast_results = pd.DataFrame({
        'Reporting Period': ['Q4', 'Q1', 'Q2'],
        'Prev Year': [
            s_actual_productRevenue[s_actual_productRevenue.index=='2016 Q4'].values[0] if (s_actual_productRevenue.index=='2016 Q4').sum() else 0,
            s_actual_productRevenue[s_actual_productRevenue.index=='2017 Q1'].values[0] if (s_actual_productRevenue.index=='2017 Q1').sum() else 0,
            s_actual_productRevenue[s_actual_productRevenue.index=='2017 Q2'].values[0] if (s_actual_productRevenue.index=='2017 Q2').sum() else 0,
        ],
        'Forecasted': [
            s_transactions[s_transactions.index=='2017 Q4'].values[0] + outlier_sum/4,
            s_transactions[s_transactions.index=='2018 Q1'].values[0] + outlier_sum/4,
            s_transactions[s_transactions.index=='2018 Q2'].values[0] + outlier_sum/4,
        ],
    })
    forecast_results['YoY (%)'] = ((forecast_results['Forecasted'] - forecast_results['Prev Year'])
                                    / forecast_results['Prev Year'] * 100).apply(lambda x: '{:+.0f}%'.format(x))
    forecast_results['Prev Year'] = forecast_results['Prev Year'].apply(lambda x: '${:,.0f}'.format(x))
    forecast_results['Forecasted'] = forecast_results['Forecasted'].apply(lambda x: '${:,.0f}'.format(x))
    forecast_results.set_index('Reporting Period', inplace=True)
    forecast_results.to_csv('../../results/tables/sales_forecast_{}.csv'.format(slugify(product_name)))
    
    return forecast, forecast_results, fig

















In [47]:



    


forecast, forecast_results, fig = product_forcast(
    df,
    'Google Men\'s  Zip Hoodie',
    yearly_seasonality=8,
)
plt.show()
forecast_results


























Ignoring 8 outlier points ($20,158.726427)
Found 460 product transactions
Saving figure to ../../results/figures/sales_forecast_google-men-s-zip-hoodie.png
Generating quarterly forecasts





























Out[47]:











  
    

      
      Prev Year
      Forecasted
      YoY (%)
    

    

      Reporting Period
      
      
      
    

  
  
    

      Q4
      $8,709
      $5,952
      -32%
    

    

      Q1
      $7,208
      $4,933
      -32%
    

    

      Q2
      $4,441
      $1,714
      -61%
    

  


























Run for each of the top 10 Top selling products













In [48]:



    


for product in df.v2ProductName.value_counts(ascending=False).head(10).index.tolist():
    print('-'*20)
    print(Fore.RED + product + Style.RESET_ALL)
    forecast, forecast_results, fig = product_forcast(
        df,
        product,
        yearly_seasonality=8,
    )
    plt.show()
    display(forecast_results)
    print()


























--------------------
Google Sunglasses
Ignoring 16 outlier points ($7,733.895237)
Found 1348 product transactions
Saving figure to ../../results/figures/sales_forecast_google-sunglasses.png
Generating quarterly forecasts








































  
    

      
      Prev Year
      Forecasted
      YoY (%)
    

    

      Reporting Period
      
      
      
    

  
  
    

      Q4
      $5,561
      $7,224
      +30%
    

    

      Q1
      $4,979
      $6,786
      +36%
    

    

      Q2
      $8,549
      $10,545
      +23%
    

  




















--------------------
Google Laptop and Cell Phone Stickers
Ignoring 25 outlier points ($4,510.63473)
Found 763 product transactions
Saving figure to ../../results/figures/sales_forecast_google-laptop-and-cell-phone-stickers.png
Generating quarterly forecasts








































  
    

      
      Prev Year
      Forecasted
      YoY (%)
    

    

      Reporting Period
      
      
      
    

  
  
    

      Q4
      $2,545
      $1,725
      -32%
    

    

      Q1
      $1,845
      $938
      -49%
    

    

      Q2
      $1,731
      $904
      -48%
    

  




















--------------------
Google 22 oz Water Bottle
Ignoring 18 outlier points ($12,721.109955)
Found 766 product transactions
Saving figure to ../../results/figures/sales_forecast_google-22-oz-water-bottle.png
Generating quarterly forecasts








































  
    

      
      Prev Year
      Forecasted
      YoY (%)
    

    

      Reporting Period
      
      
      
    

  
  
    

      Q4
      $11,051
      $-712
      -106%
    

    

      Q1
      $5,580
      $-6,469
      -216%
    

    

      Q2
      $7,215
      $-4,928
      -168%
    

  




















--------------------
Google Men's 100% Cotton Short Sleeve Hero Tee White
Ignoring 14 outlier points ($6,400.144081)
Found 746 product transactions
Saving figure to ../../results/figures/sales_forecast_google-men-s-100-cotton-short-sleeve-hero-tee-white.png
Generating quarterly forecasts








































  
    

      
      Prev Year
      Forecasted
      YoY (%)
    

    

      Reporting Period
      
      
      
    

  
  
    

      Q4
      $6,543
      $6,561
      +0%
    

    

      Q1
      $5,421
      $5,246
      -3%
    

    

      Q2
      $3,261
      $3,617
      +11%
    

  




















--------------------
Google Men's 100% Cotton Short Sleeve Hero Tee Black
Ignoring 12 outlier points ($7,077.371756)
Found 703 product transactions
Saving figure to ../../results/figures/sales_forecast_google-men-s-100-cotton-short-sleeve-hero-tee-black.png
Generating quarterly forecasts








































  
    

      
      Prev Year
      Forecasted
      YoY (%)
    

    

      Reporting Period
      
      
      
    

  
  
    

      Q4
      $7,706
      $8,730
      +13%
    

    

      Q1
      $3,948
      $5,775
      +46%
    

    

      Q2
      $4,266
      $5,954
      +40%
    

  




















--------------------
Google Men's 100% Cotton Short Sleeve Hero Tee Navy
Ignoring 7 outlier points ($1,606.964444)
Found 554 product transactions
Saving figure to ../../results/figures/sales_forecast_google-men-s-100-cotton-short-sleeve-hero-tee-navy.png
Generating quarterly forecasts








































  
    

      
      Prev Year
      Forecasted
      YoY (%)
    

    

      Reporting Period
      
      
      
    

  
  
    

      Q4
      $3,671
      $5,082
      +38%
    

    

      Q1
      $1,890
      $3,576
      +89%
    

    

      Q2
      $2,110
      $3,799
      +80%
    

  




















--------------------
Google Men's  Zip Hoodie
Ignoring 8 outlier points ($20,158.726427)
Found 460 product transactions
Saving figure to ../../results/figures/sales_forecast_google-men-s-zip-hoodie.png
Generating quarterly forecasts








































  
    

      
      Prev Year
      Forecasted
      YoY (%)
    

    

      Reporting Period
      
      
      
    

  
  
    

      Q4
      $8,709
      $5,952
      -32%
    

    

      Q1
      $7,208
      $4,933
      -32%
    

    

      Q2
      $4,441
      $1,714
      -61%
    

  




















--------------------
Recycled Paper Journal Set
Ignoring 21 outlier points ($7,586.310712)
Found 410 product transactions
Saving figure to ../../results/figures/sales_forecast_recycled-paper-journal-set.png
Generating quarterly forecasts








































  
    

      
      Prev Year
      Forecasted
      YoY (%)
    

    

      Reporting Period
      
      
      
    

  
  
    

      Q4
      $6,731
      $7,274
      +8%
    

    

      Q1
      $4,153
      $4,379
      +5%
    

    

      Q2
      $6,668
      $7,094
      +6%
    

  




















INFO:fbprophet:n_changepoints greater than number of observations.Using 15.0.

















--------------------
BLM Sweatshirt
Ignoring 13 outlier points ($1,624.856666)
Found 392 product transactions
Saving figure to ../../results/figures/sales_forecast_blm-sweatshirt.png
Generating quarterly forecasts








































  
    

      
      Prev Year
      Forecasted
      YoY (%)
    

    

      Reporting Period
      
      
      
    

  
  
    

      Q4
      $0
      $-42,853
      -inf%
    

    

      Q1
      $3,363
      $-81,210
      -2515%
    

    

      Q2
      $12,085
      $-102,801
      -951%
    

  




















--------------------
Engraved Ceramic Google Mug
Ignoring 13 outlier points ($7,495.080555)
Found 387 product transactions
Saving figure to ../../results/figures/sales_forecast_engraved-ceramic-google-mug.png
Generating quarterly forecasts








































  
    

      
      Prev Year
      Forecasted
      YoY (%)
    

    

      Reporting Period
      
      
      
    

  
  
    

      Q4
      $5,076
      $4,001
      -21%
    

    

      Q1
      $2,789
      $1,974
      -29%
    

    

      Q2
      $526
      $292
      -44%
    

  






































In [117]:



    


from IPython.display import HTML
HTML('<style>div.text_cell_render{font-size:130%;padding-top:50px;padding-bottom:50px}</style>')




















Out[117]: